Neural substrates of predicting anhedonia symptoms in major depressive disorder via connectome‐based modeling

Tingyu Yang, Yangpan Ou, Huabing Li, Feng Liu, Ping Li, Guangrong Xie, Jingping Zhao, Xilong Cui, Wenbin GuoCNS 2024 · DOI

TL;DR

This study leverages connectome-based predictive modeling (CPM) and support vector machines (SVM) to predict anhedonia symptoms in Major Depressive Disorder (MDD) using resting-state fMRI data. They identified specific brain network patterns associated with anhedonia and successfully differentiated between melancholic and non-melancholic MDD subtypes, offering potential biomarkers for clinical guidance.

What Problem Does This Paper Address?

Anhedonia, the inability to experience pleasure or loss of interest, is a core symptom of Major Depressive Disorder (MDD) and is linked to poorer prognosis, increased relapse risk, and poorer treatment response. Understanding the neural mechanisms underlying anhedonia is crucial for developing more effective diagnostic and treatment strategies for MDD. While previous studies have explored functional connectivity (FC) abnormalities in MDD patients with anhedonia, findings have been inconsistent, with abnormal FC patterns distributed across various brain regions. Furthermore, MDD is increasingly recognized as a heterogeneous disorder, with melancholic and non-melancholic subtypes potentially representing distinct biological entities. This paper aims to address these gaps by developing a robust predictive model for anhedonia symptoms using whole-brain FC and investigating whether specific FC patterns can differentiate between these two MDD subtypes.

Methods

The study employed two primary machine learning techniques:

  • Connectome-Based Predictive Modeling (CPM): This data-driven approach uses whole-brain functional connectivity (FC) data to build predictive models. In this study, CPM was used to predict anhedonia symptom severity. The process involved identifying "high-degree nodes" (regions with many connections) and then computing their FC with other nodes.
  • Support Vector Machines (SVM): SVM is a classification algorithm used to distinguish between groups. Here, it was applied to classify MDD patients into melancholic and non-melancholic subtypes based on their FC patterns. The study utilized a five-fold cross-validation strategy for training and testing, optimizing SVM parameters (c and g) using grid search.

The study also incorporated node-based FC analysis to compare specific connectivity patterns between the two MDD subtypes.

Datasets

The study utilized resting-state functional magnetic resonance imaging (fMRI) data from:

  • 64 patients diagnosed with MDD (31 melancholic, 28 non-melancholic).
  • 32 healthy controls (HCs).

Data were collected from the Second Xiangya Hospital of Central South University.

Key Insights and Ideas

  • CPM as a Powerful Predictor: The study demonstrates that CPM can effectively predict anhedonia symptoms in MDD patients, suggesting that whole-brain FC patterns hold significant predictive power for this symptom.
  • Distinct Neural Substrates for MDD Subtypes: The research highlights that melancholic and non-melancholic MDD patients exhibit distinct FC patterns, particularly between the left cingulate gyrus (CG_L) and the right parahippocampal gyrus (PhG_R). This suggests that these subtypes may have different underlying neurobiological mechanisms.
  • PhG_R and CG_L as Potential Biomarkers: The specific FC between PhG_R and CG_L emerged as a key differentiator between melancholic and non-melancholic MDD patients, showing high accuracy in classification. This connection could serve as a valuable neurobiological marker for clinical differentiation.
  • FC-Anhedonia Correlation: A higher FC value between PhG_R and CG_L was associated with fewer anhedonia symptoms, implying that disruptions in this specific connectivity might contribute to the experience of anhedonia.
  • Intrinsic vs. Symptomatic Differences: The findings suggest that the observed FC differences between MDD subtypes might be intrinsic neural substrates rather than solely a consequence of anhedonia symptoms, especially since no significant differences in depression severity were found between the groups.

Key Results

  • Anhedonia Prediction: CPM successfully predicted anhedonia symptoms in MDD patients with a positive network correlation of r=0.4719 (p<0.0020).
  • Subtype Classification: SVM analysis using FC between PhG_R and CG_L achieved a classification accuracy of 81.36% for differentiating melancholic from non-melancholic MDD patients, with a sensitivity of 93.54% and specificity of 67.86%.
  • Specific FC Differences: Melancholic MDD patients showed decreased FC between the left cingulate gyrus (CG_L) and the right parahippocampal gyrus (PhG_R) compared to non-melancholic MDD patients (p-bonferroni =0.0303).
  • Correlation with Clinical Variables: FC between PhG_R and CG_L was positively correlated with TEPS consummatory scores (p=0.0016) and TEPS total scores (p=0.0200) in MDD patients, after controlling for covariates.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: The relatively small sample size in each group may limit the generalizability of the findings.
  • Independent Validation: The predictive model has not been validated on an independent sample, which is crucial for assessing its robustness.
  • Cross-Sectional Design: The study's cross-sectional nature prevents the establishment of a direct causal relationship between FC changes and anhedonia.

Future work should focus on:

  • Longitudinal Studies: Conducting longitudinal studies to investigate the stability of FC in predicting anhedonia and classifying MDD subtypes.
  • Larger Sample Sizes: Recruiting larger cohorts to improve the generalizability of the findings.
  • Independent Validation: Validating the developed models on independent datasets.
  • Exploring Other Neuroimaging Modalities: Integrating other neuroimaging techniques (e.g., structural MRI) to gain a more comprehensive understanding of the neural underpinnings of anhedonia in MDD.

Summary updated

Abstract

This study successfully established a network model for predicting anhedonia symptoms in MDD based on FC, as well as a classification model to differentiate between melancholic and non-melancholic MDD patients. These findings provide guidance for clinical treatment.
BibTeX
@inproceedings{yang2024neural,
  author = {Tingyu Yang and Yangpan Ou and Huabing Li and Feng Liu and Ping Li and Guangrong Xie and Jingping Zhao and Xilong Cui and Wenbin Guo},
  title = {{Neural substrates of predicting anhedonia symptoms in major depressive disorder via connectome‐based modeling}},
  booktitle = {CNS},
  year = {2024},
  doi = {10.1111/cns.14871},
}
Figure